Background:

Despite recommendations from the U.S. Preventive Services Task Force (USPSTF) regarding the age at which to begin and end cervical cancer screening, as well as the interval at which to conduct screening, there is limited direct evidence beyond that inferred from epidemiologic and natural history studies to support these recommendations. In addition, concerns about the poor sensitivity (approximately 50 percent) of cytology-based screening have led to the development of new tests with potentially improved sensitivity for the detection of cervical intraepithelial neoplasia (CIN) grades 2 and 3. Although there is widespread use of these tests—including the Hybrid Capture 2 high-risk human papillomavirus (HPV) deoxyribonucleic acid (DNA) test—the USPSTF has, to date, not recommended their use due to a lack of definitive evidence regarding their performance in screening. The availability of new data, including data from randomized controlled trials, suggests a need to re-evaluate the previous recommendations. Simulation modeling can provide additional guidance on the risks, benefits, and resources associated with different screening test strategies, as well as the trade-offs involved in varying the age at which to begin and end screening.

Purpose:

A decision model was used to address two specific aims: 1) How many colposcopies per life-year gained are associated with each of the different ages for beginning screening for cervical cancer (varying in 1 year increments from ages 15 to 25 years)? and 2) How many colposcopies per life-year gained are associated with cervical cancer screening strategies that use HPV DNA testing in conjunction with cytology, compared to strategies based on cytology only?

In addition, as a sub-aim of Specific Aim 1, the age at which to end screening for cervical cancer in women who have previously been screened every 3 years prior to age 65 years or who have never been screened was also examined.

Methods:

The model used for the analysis (the Duke Cervical Cancer model) was developed as part of a previous evidence report prepared for the Agency for Healthcare Research and Quality. The model describes the natural history of HPV infection, including progression to CIN2-3 and cancer, as well as the impact of screening and treatment on the prevention of disease progression in a cohort of unvaccinated girls who are followed until either death or age 100 years. Test characteristics for the different screening tests are primarily based on a companion evidence report prepared by the Oregon Evidence-based Practice Center. For each question, outcomes presented include (per 1,000 women): false-positive test results, colposcopies performed, cases of CIN2-3, cases of cervical cancer, and cervical cancer deaths. The main outcome is colposcopies per (undiscounted) life-year. This outcome, which is not based on cost, was chosen by the USPSTF for the primary analysis as a metric that best represents a reasonable trade-off between the burden and benefits of screening. Strategies are compared using incremental ratios. Strategies that are associated with 1) more colposcopies but less effectiveness or 2) fewer colposcopies but higher incremental colposcopies per life-year than an adjacent strategy are considered to be dominated and are eliminated from consideration for this analysis. The remaining strategies (after this elimination process) lie on an “efficiency” frontier (although efficiency in this context is measured using colposcopies per life-year instead of cost per life-year) and, as such, may represent a reasonable trade-off between the burden and benefits of screening.

Results:

An analysis of the age at which to begin screening shows that screening with cytology in the teens is associated with a high number of false-positive test results and few detected cases of cancer. Analyses using the metric of colposcopies per life-year suggest that screening less frequently than annually beginning in the twenties might provide a reasonable trade-off between the burden and benefits of screening. However, since American Society for Colposcopy and Cervical Pathology guidelines recommend rescreening instead of immediate referral to colposcopy for women younger than age 21 years, colposcopies per life-year may underestimate the burden of screening in this age group. A sensitivity analysis that uses number of screening cytology tests instead of colposcopies as the metric of interest also identifies screening strategies that begin at later ages, including the USPSTF’s current recommended strategy of beginning screening no later than age 21 years, and conducted at least every 3 years, as strategies that may better represent a reasonable trade-off between the burden and benefits of screening.

In terms of the age at which to end screening, among women who have never been screened prior to age 65 years, strategies associated with screening every 2 to 5 years and ending in the 70s are identified as representing a reasonable trade-off between the burden and benefits of screening. Beyond this decade, the gains in life expectancy are small compared to the number of colposcopies performed. Among women who have been screened every 3 years prior to age 65 years, the incremental colposcopies per life-year gained associated with any further screening are high for all strategies due to the smaller gains in life expectancy. These findings are robust across a range of sensitivity analyses.

Analyses comparing cytology with and without HPV testing show that identifying co-testing (cytology and HPV, with screening every 3 years assumed for women with HPV negative and cytology normal results) as an efficient strategy depends on how the burden of screening is quantified. If colposcopies per life-year is used as the outcome, co-testing strategies are identified as efficient. However, if screening and triage tests are used to quantify burden, cytology-only strategies are identified as more efficient than co-testing strategies. In sensitivity analyses, a strategy of HPV testing followed by cytology for high-risk HPV positive women, with referral to colposcopy if both tests are abnormal, is consistently identified as efficient, regardless of whether colposcopies or tests (screening and triage) are used to quantify burden.

Conclusions:

This decision analysis supports current recommendations regarding the age at which to begin and end screening. A strategy of co-testing with cytology and HPV (and screening every 3 years for women with dually negative results) is identified as efficient compared to cytology if colposcopies are used to quantify burden. However, if tests are used to quantify burden, cytology-only strategies are identified as efficient compared to co-testing. A sensitivity analysis suggests that a strategy of HPV followed by cytology (for women with HPV positive test results) warrants further study.